Microvascular invasion is an important adverse histopathologic feature in hepatocellular carcinoma, yet it cannot be directly established before treatment. In a large multicentre study, MAPUSE used routine contrast-enhanced…

PregMedNet offers one of the largest systematic maps of maternal medication exposure and neonatal complications to date. Across more than 1.19 million mother–baby dyads, respiratory outcomes emerged as…

Recognizing the Glissonean pedicle and hepatic veins during laparoscopic liver resection is a demanding visual task. In a multicenter retrospective study, the video-aware AI model Vivim delivered leading…

Routine clinical imaging data may be a powerful substrate for medical AI. In this study, a self-supervised 3D model trained on more than 5 million CT and MRI…

A new AI system, RVOTEP, was developed to predict longitudinal anti-VEGF treatment effects in retinal vein occlusion using OCT and fundus images. In internal and external testing, it…

Conventional sleep staging compresses a night of brain activity into five labels. This study asks whether self-supervised EEG models can preserve health-relevant structure inside those labels.

IBOTSpine achieved the strongest overall results among eight tested architectures for classifying normal, scoliosis, and spondylolisthesis X-rays. Its combination of feature compression and optimal-transport alignment was designed to…

Recent Hematology coverage shows AI shifting toward explainable risk tools, dynamic prognosis, transplant support, new sensing, morphology assistance, and target discovery, with validation now the central test.

Transcript INTRO Welcome to Medical AI Digest Podcast. Today, a paper from Nature communications. MAIN Today we’re looking at a computational study on neuroblastoma, focused on a model…

A previously trained AI system estimated gestational age from blind sweep ultrasound videos collected by novice operators with accuracy noninferior to standard expert sonographer-based clinical estimates.